The Artificial Intelligence Podcast

China has unveiled a new AI chatbot trained on President Xi Jinping's political ideology, a significant step in spreading his ideas and a reminder of the ideological parameters AI models in China must operate within. The chatbot was trained on seven databases, including one dedicated to Xi Jinping Thought, a doctrine promoting socialism with Chinese characteristics. Its creation is part of a broader effort to promote Xi's ideologies, already integrated into education and available through an app. The chatbot can generate reports, summarize information, and provide translations, highlighting China's ambitions to lead in AI by 2030.

A $1.5 billion deal between Microsoft and G42, a UAE-based AI company, has sparked concerns among US lawmakers who fear sensitive AI technology could be exposed to Chinese espionage. G42 has ties to Chinese tech and surveillance companies, but promises to adhere to new safeguards and stop using Chinese equipment. Lawmakers worry the deal could still give G42 access to advanced semiconductors the US wants to protect from China. The deal raises questions about balancing tech expansion with protecting sensitive technology, with some lawmakers demanding more transparency and scrutiny.

Colorado has become the first state to pass legislation aimed at addressing bias in artificial intelligence programs. The law requires companies using AI to assess the risk of discrimination and inform customers when AI is used in consequential decisions. It also mandates oversight programs and reporting of discrimination instances to the state attorney general. This development is significant as AI algorithms are increasingly used in decisions related to hiring, housing, and medical care, and can perpetuate discrimination due to biases in historical training data.

Google's new artificial intelligence tool is facing criticism for providing incorrect or misleading responses to user queries. Launched as an experimental feature, it generates an "AI overview" response based on web sources and is placed at the top of some search results. However, users have taken to social media to share examples of wrong or misleading results, including false information about elephants' feet and Barack Obama's religion. Google attributes the errors to uncommon queries and users trying to trip up the technology, but the controversy raises concerns about the reliability of artificial intelligence in providing accurate information.

Amazon is upgrading its Alexa voice assistant with generative artificial intelligence technology to compete with Google and Open AI's chatbots. The new Alexa, launching later this year, will offer a more conversational experience and come with a monthly subscription fee. This move marks a significant shift from Amazon's current business model and is seen as essential to keeping up with rapid advancements in AI technology. Amazon's large installed user base of over 500 million Alexa-enabled devices gives it an advantage, but it faces challenges in finding AI engineering talent and offsetting the cost of using generative AI.

The CIA is leveraging generative artificial intelligence to enhance operations and stay ahead of adversaries. The agency's first chief technology officer, Nand Mulchandani, is leading the effort to develop AI applications that augment human capabilities, rather than replace them. The CIA is creating its own generative AI platform, similar to Chat GPT, to help analysts process vast amounts of data and spark creative thinking. However, the agency is cautious about the technology's limitations and potential biases, recognizing the need for human judgment to weigh the importance of incoming information.

Google has unveiled its sixth-generation Tensor Processing Unit, called Trillium, which boasts significant improvements in performance and energy efficiency compared to its predecessors. The new chip offers 4.7 times more peak compute performance, doubled high-bandwidth memory, and enhanced inter-chip communication, making it capable of handling AI tasks faster and more efficiently. It's also 67% more energy-efficient than the previous TPU v5e, allowing it to run for longer periods without excessive power consumption. The Trillium chip will be available in Google Cloud, although pricing and availability details have not been disclosed.

This week, South Korea is hosting a significant gathering focused on the risks and regulation of artificial intelligence, building on the work started at last year's inaugural AI safety conference in Britain. The Seoul summit aims to mitigate the threats posed by cutting-edge artificial intelligence systems, including algorithmic bias and existential risks to humanity. Delegates from over two dozen countries will attend, including the US and China, to discuss how to address the risks of AI and set shared standards to manage it. The meeting will feature updates from leading AI companies and digital ministers sharing best practices to protect society from AI's negative impacts.

A new federal assessment warns that artificial intelligence programs pose a significant threat to the democratic system of elections, particularly with the 2024 election cycle approaching. Next-generation technologies can be exploited to influence and sow discord in elections by creating or sharing altered or deepfaked pictures, videos, or audio clips. The Department of Homeland Security analysis outlines how generative AI tools can be used to confuse or overwhelm voters and election staff. Experts caution that authorities must be prepared to defend against AI-disseminated fake news and educate the public to counteract misinformation.